Anthropic's Fable 5 AI Is Back Online After a Three-Week Government Ban
The U.S. Commerce Department lifted its export restrictions on Anthropic's Fable 5 model Tuesday, after weeks of closed-door negotiations over whether the AI could be weaponised by bad actors.
Key points
- Anthropic's Fable 5 AI model was taken offline on June 12, 2025, after senior U.S. government officials said it posed severe cybersecurity risks.
- Commerce Secretary Howard Lutnick notified Anthropic co-founder Tom Brown by letter on Tuesday, June 24, 2025, that the export control restrictions — rules preventing foreign nationals from accessing the model — were lifted.
- Anthropic's more powerful Mythos 5 model returned to roughly 100 vetted organisations on Friday, June 21, 2025, but remains unavailable to the general public.
- Anthropic agreed in writing to share information about malicious activity with the U.S. government and to collaborate on safety standards for future models.
- Experts at the Commerce Department's Center for AI Standards and Innovation tested and signed off on new safeguards Anthropic built to address the original concerns.
For three weeks, one of the most capable AI systems available to the public simply didn't exist — at least not for anyone who tried to log on. Anthropic's Fable 5, part of its Claude family of AI models, went dark on June 12 after the U.S. Commerce Department ordered it offline. Access returned Wednesday.
The reason for the ban: senior administration officials worried that Fable 5's guardrails — the built-in rules designed to stop the AI from helping with dangerous cybersecurity or biology tasks — could be bypassed by a determined user. A bypass like that, officials feared, could hand a foreign adversary a surprisingly capable tool for causing harm.
How did the government even have the power to do this?
The Commerce Department used export controls — laws that let the U.S. government restrict which foreign countries or individuals can access sensitive American technology — to force Anthropic's hand. Because complying with the order meant verifying every single user's nationality, Anthropic simply took both Fable 5 and the even more powerful Mythos 5 offline entirely on June 12, affecting customers worldwide.
Not everyone agreed the risk was real. Cybersecurity experts were divided on whether the guardrail bypass was a serious threat or a theoretical edge case. Sam Altman, CEO of Anthropic competitor OpenAI, called the government's new habit of approving AI releases in stages "bad news." Industry leaders fretted openly that the ad hoc ban would hand China a window to close the gap on American AI.
Anthropics response was fast. The company sent a team of its top researchers to Washington, D.C., to work out a fix directly with government officials. Those talks produced two things: a technical solution — a new, more targeted method of blocking the specific workaround that had alarmed officials — and a formal agreement committing Anthropic to keep the government informed of any malicious activity it detects.
First reported by NBC News, the letter from Secretary Lutnick to Tom Brown confirmed the deal and framed it as a win for "America's leadership in AI."
Fable 5 is now back for the general public. Mythos 5 remains limited to roughly 100 U.S.-based organisations working on cybersecurity and critical infrastructure, while Anthropic says it will push to restore broader access.
This episode is a signal. For years, Washington mostly watched the AI industry from a distance. The forced pause on Fable 5 and Mythos 5 — however contested — shows that governments now feel ready to pull the plug when they think the stakes are high enough.
If you use Fable 5 for work, the practical step right now is simple: check whether your organisation's access has fully restored, and flag anything that looks unusual in the model's responses to your IT or security team.
